Statistical Analysis for Next-Generation Sequencing data in Family-based designs

Abstract

Genome-wide association studies (GWAS) typically involve examination of 100,000 to more than 1,000,000 genetic variants, such as single nucleotide polymorphisms (SNPs), in different individuals to identify SNPs associated with a disease. Since the conclusion of the Human Genome Project, this project elucidated understand human genetic variation and paved the way for the GWAS approach. GWAS have successfully identified thousands of common genetic loci associated with many phenotypes. Despite the success of GWAS, the variants identified by these studies have often explained only a small fraction of heritability for most phenotypes, and this observation underscored the importance of studying rare or less common variants. Contrary to the traditional GWAS approach, single variant association analysis with rare variants has difficulties with detection of causal variants. To overcome the issue with statistical power in rare variant association studies, researchers have recently developed statistical methods for testing rare variants in a population-based design. Because individuals in a family are genetically more homogenous than unrelated individuals, family-based designs can play an important role in rare-variant studies. Despite the importance of rare variant analysis for the family-based design, only a few statistical methods for family-based rare-variant association analysis are available. Furthermore, even though many genes on the X chromosome are related to human complex diseases, few significantly associated rare variants have been identified on the X chromosome for complex traits. In this study, we propose a family-based rare-variant association test (FARVAT) and a family-based rare-variant association test for X-linked genes (FARVATX). FARVAT is based on quasi-likelihood, and is statistically and computationally efficient for the family-based design. We considered that families were ascertained with the disease status of family members, and we calculated the genetic relationship matrix for the proposed method

this matrix ensured robustness in the presence of population substructure. Depending on the choice of a working matrix, FARVAT could be a burden-type or a variance component-type method, and could be extended to the optimal-type method. In the analysis of the X chromosome, FARVATX can accommodate random X chromosome inactivation (XCI), escaped XCI (E-XCI), and skewed XCI (S-XCI). FARVATX is computationally efficient and can complete X-linked analyses within a few hours. With extensive simulation studies under various scenarios, we compared the proposed methods with the existing ones, and the results showed that the proposed methods are the more powerful in terms of simulation settings. We also applied FARVAT and FARVATX to schizophrenia data and chronic obstructive pulmonary disease (COPD) data, respectively. The proposed methods may help researchers identify additional X-linked rare variants associated with complex traits, thereby leading to a better understanding of the underlying biological processes associated with X-linked genes.